Discussion in 'Videocards - AMD Radeon Drivers Section' started by Havel, Nov 24, 2015.
Mine barely hit 70 on either card so for me and my cards so yes this was abnormal.
I guess I should mention it was climbing I noticed it 103 and by the time I was able to get into OD, Accept OD etc it was at 105. Game had been closed as well.
you missed my point.
yes sounds like there is an issue if those readings are correct.
My point was your cards wont melt, 105 is within spec, and they are protected.
It's abnormal if it's caused by the drivers and not usually that hot
i doubt its caused by the drivers.
They sound like erroneous reading anyway.
power is a calculation (I X E), its not measured.
If your voltage is off way too high, so will be power.
This might be the cause.
would have throttled if that was the issue.
several people have already tested it.
Plus hes using the hotfix, that issue has been resolved.
still sound like bogus readings to me.
id try repeating it with aida64 and see what you get.
I can say I had set my fans to run at 100% and they were running at 3700RPM and 3200 RPM and I know due to the noise they will make.
Anyhow the point I was trying to make is things may or may not be working completly yet. Could be a number of things that caused this, but for me this is not nomral behavor.
I use HIS iTurbo to OC since I do not need all the other things with AB but thought I would try them out with the new fixed crimson driver.
Most of us here have expensive parts, so monitoring things during demanding task would not be be a bad idea for the time being.
I do not even enable OD, not messed with it in years, but had to undo
MY GPU temp never got above 70 degrees just that one VRM
And yes it could of been just a fluke of something.
I'm beginning to suspect all Crimson issues are related to incorrect voltage regulation.
I too was skeptical of people claiming the driver fried their cards as they should throttle.
However if the temp spikes were caused by overvolting the core by more than twice the nominal, surely the core could fry before the throttling gets to trigger!
Also on the hotfix thread someone said he could fix the artefacts by forcing constant voltage in AB. I got similar artefacting issues if I overclocked my GPU too high without enough voltage.
In turn the clockspeed variations could be due to undervolting. The Power Limit is a temperature and current capability threshold AFAIK.
Power is defined as current times voltage. If the voltage drops, then current must be higher to achieve the same power level. This could cause the current limit to trigger and thus the card to downclock.
Yeah the problem can't be caused by drivers.Impossible, never occured before...Lightning 290X fan problem possibly was a covert Nvidia operation.
The real problem was identified by AMD:" low fan speed reports" SPECIFICALLY "on select GPUs".
Thats EXACTLY the problem: THE REPORTS!
Who dare to suggest the possibility of a problem in the drivers.
Must be some Nvidia fanboys!
3.89 / 1.22 = ~ 3.19, that's 319% - being that a decent overclock on components could not even be 10%, if the card still works, then for sure the readings were wrong for some reason or another.
Has anyone found a solution for the gpu clock bouncing up and down ?...
For info I haven't activated overdrive, tried to put fps target at 200fps (lol) and disabled vsync (don't have freesync)
The only way I can stop the clock from jumping in the witcher 3 is by putting everything on ultra (even hairworks), in TW3 if I don't activate hairworks the game runs worse coz it goes down to like 700-900mhz -_-, though it still does jump a bit with everything on ultra but not as drastically from 1000 ish to 1050.
and on games like Assetto Corsa/Project cars, running at a rock solid 1050mhz no problems, and just brings down gpu usage as expected.
I think a large part of what everyone talks about clocks and them varying has to do with the actual game being played itself.
To give an example, when I play Bioshock Infinite which had a large budget I'd imagine, gpu stays pretty much pinned at 1200 Mhz without artificially locking. Lots of games are like that here. Then on the other hand, there are quite a few games that do not max GPU use or use full available clock while the CPU is not at 100% time.
I know there are different cards and I am sure the driver is not beyond critique, but the games and/or engines probably play a larger role than given credit for. I generally do not ever feel "damn the driver is being funky" or whatnot.
Slight load based variation is ofc normal. Clocks jumping from 100 - 1000 or whatnot based on what people have reported within the same game is not.
I don't think I've ever seen larger than 50 - 100 MHz variations while monitoring. The "base" clock should be set according to power state (load) and then just kind of fine tune from there.
Meaning that for example my card should never go below the 3D Perf power state base setting, set in vBIOS when running a 3D application (IIRC something like 850 MHz).
Large variations could mean the card is switching power states all the time. Or the undervoltage thing. Or something else completely. Or the combination of all of the above
It looks like it's trying to target 60fps, even though fps target is not activated, by varying the voltage/clock speed, and it's doing it badly. A bit like what you would expect a nano to do.
There is something else i've noticed aswell with crimson, the screen corruption problem had almost gone, I got it once since release whereas I get at the very least one a day (most of the time more) on all the other drivers. And from what i've understood it's something to do with the zero core power state. So I think they have changed something in the drivers to do with voltage regulation, which is messing with everything.
And the funny thing is, is that the clock goes down in the more demanding areas once you leave those areas it goes back up to full speed.
Also it seems to be happening with games that AMD have decided should have shader cache activated, GTA V and tw3 in my case.
But TW3 does run without stutter with the crimson drivers which is good (though I didn't have to much stutter in tw3) but now it slows down instead -_-
Did you play at 1080p?
Real 4K or VSR 4K (in your case in mine 1800p VSR) could help to put more load on GPU and avoid clocks goes down.
At 1080p there is not enough load on GPU in some game situations where AMD drivers DX11 API overhead low performance bottleneck the CPU and the GPU must wait but now in new Crimson drivers we have the "improved" power saving downclocking GPU also.
Increasing the res could help with GPU load and Crimson "improvements".
More or less every thing, real 4k vsr 4K(for 60hz) and vsr 1440p and of course 1080p.
The clocks jumped everywhere. The only way I could get a stable FPS was with everything ultra and HW activated at 1080p, this was never a problem previously on other drivers, it shouldn't be a problem now.
On TW3 I used to play in vsr 4K with custom settings, (high/ultra no hairworks) and got 60fps with these drivers I was getting around 40fps in demanding areas due to the low gpu clock.
And then the problem bringing up the resolution on these new drivers is that there is no sweet spot, that sweet spot I have it with the 15.11.1 ccc.
Do you create a D3D profile within AB and disabled ULPS?
I don't like the clocks throttling, i even use force constant voltage.
No D3D profile, ULPS disabled or enabled no difference, which is logic as i'm not in crossfire, even uninstalled AB, didn't change a thing. It's definitely the drivers there is no question about that, the question is how to avoid it or when will it be fixed.
The aggressive clock throttling is caused by Crimson but i managed to avoid this problem keeping my clocks stable using AB profiles.
About drivers fix date, this is a very difficult question. :3eyes:
AMD first need to acknowledge the problem and not only the fan problem reports.